linux-kernel - Re: [PATCH v3 6/6] asus-wmi: Support the hardware GPU MUX on some laptops

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Fri, 12 Aug 2022 10:44:54 +0200
From:   Hans de Goede <hdegoede@...hat.com>
To:     Thomas Weißschuh <thomas@...ch.de>
Cc:     Luke Jones <luke@...nes.dev>, andy.shevchenko@...il.com,
        pobrn@...tonmail.com, pavel@....cz,
        platform-driver-x86@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 6/6] asus-wmi: Support the hardware GPU MUX on some
 laptops

Hi,

On 8/12/22 10:31, Thomas Weißschuh wrote:
> Hi,
> 
> On 2022-08-12 09:59+0200, Hans de Goede wrote:
>> Date: Fri, 12 Aug 2022 09:59:29 +0200
>> From: Hans de Goede <hdegoede@...hat.com>
>> To: Luke Jones <luke@...nes.dev>
>> Cc: andy.shevchenko@...il.com, pobrn@...tonmail.com, pavel@....cz,
>>  platform-driver-x86@...r.kernel.org, linux-kernel@...r.kernel.org
>> Subject: Re: [PATCH v3 6/6] asus-wmi: Support the hardware GPU MUX on some
>>  laptops
>> User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
>>  Thunderbird/91.12.0
>>
>> Hi,
>>
>> On 8/12/22 00:01, Luke Jones wrote:
>>> Hi Hans,
>>>
>>> On Thu, 2022-08-11 at 15:53 +0200, Hans de Goede wrote:
>>>> Hi,
>>>>
>>>> On 8/9/22 04:50, Luke D. Jones wrote:
>>>>> Support the hardware GPU MUX switch available on some models. This
>>>>> switch can toggle the MUX between:
>>>>>
>>>>> - 0, Dedicated mode
>>>>> - 1, Optimus mode
>>>>>
>>>>> Optimus mode is the regular iGPU + dGPU available, while dedicated
>>>>> mode switches the system to have only the dGPU available.
>>>>>
>>>>> Signed-off-by: Luke D. Jones <luke@...nes.dev>
>>>>
>>>> I see that you have replied to this that it needs more work.
>>>>
>>>> Besides it needing more work, ideally this should hook into
>>>> the existing vga-switcheroo mechanism for this. Can you take
>>>> a look at that please?
>>>>
>>>> I think this might be the first non GPU driver doing vga-
>>>> switcheroo stuff. So this may be something to discuss
>>>> on the dri-devel list.
>>>
>>> I'm not sure how this would work. In typical ASUS fashion they do non-
>>> standard stuff. This switch is a basic toggle that requires a reboot to
>>> enable after writing to the ACPI method, after reboot the dGPU becomes
>>> the only visible GPU on the system and (this GPU) can not be suspended.
>>>
>>> In short: it toggles the laptop from discrete-only mode, and optimus
>>> mode, requiring a reboot to switch.
>>>
>>> From what I understand of switcheroo it is more to manage having dual
>>> (or more) GPU available during runtime, and manage the power states,
>>> offload etc.
>>
>> Right, I did not realize this requires a reboot, that would be
>> something to mention in the Documentation bits accompanying the patch.
>>
>> This is also a reason why it is good to have the docs update in
>> the same patch as adding the functionality, because the docs may
>> help with reviewing.
>>
>> Anyways I agree that if this requires a reboot then using
>> the vga switcheroo stuff is not applicable. So we can just go with
>> a simple(ish) asus-wmi sysfs attribute.
> 
> Would this not fit the existing "firmware-attributes" class?
> It even has a flag to signal that a reboot is required after an attribute has
> been changed.

Yes it sounds like a BIOS setting is being toggled from within
Linux, which would normally be done through the
"firmware-attributes" class, but all existing "firmware-attributes"
class drivers allow changing all BIOS setting not just a single
setting, so using the  "firmware-attributes" class here is not really
appropriate.

> Maybe it is overkill to use it only for this, though.

Right :)

Regards,

Hans


>>> I have a vastly improved patch for this prepared now. Because of how
>>> the actual feature works (and the above explanation) it must be under
>>> the asus-nb-wmi sysfs (next to the dgpu_disable and egpu_enable toggles
>>> which are also unusual and non-standard work-arounds of Windows
>>> issues).
>>
>> Ack, sounds good.
> 
> Thomas
>